A battery short circuit fault diagnosis method and system based on impedance spectrum

Abstract

This disclosure provides a battery short-circuit fault diagnosis method and system based on impedance spectrum, relating to the field of battery management technology. The method includes: acquiring impedance spectrum data of the battery to be diagnosed at the optimal sampling frequency band and preprocessing it; inputting the preprocessed impedance spectrum data into a trained fault diagnosis model to determine whether a short-circuit fault has occurred, and obtaining a fault diagnosis result; wherein, the optimal sampling frequency band is obtained by using the fault diagnosis model to quantify the influence of impedance data at each sampling frequency on the fault diagnosis result, obtaining the importance of frequency features, and constructing the optimal sampling frequency band based on the importance of frequency features; this invention, based on the importance of frequency features, filters the fault dataset of the fault diagnosis model, improves the information content of the fault dataset, and achieves accurate diagnosis of battery short-circuit faults even with only a small amount of battery fault data.

Description

Technical Field

[0001] This disclosure relates to the field of battery management technology, specifically to a battery short-circuit fault diagnosis method and system based on impedance spectrum. Background Technology

[0002] In practical applications of electric vehicles and energy storage systems, various internal and external faults may occur during battery operation, leading to performance problems and even serious consequences such as thermal runaway, fire, or explosion. Among these, short circuit is the most common and representative safety failure mode, and one of the most important factors causing battery failure and safety issues. A deepening short circuit can cause thermal runaway, resulting in serious safety accidents. Therefore, timely diagnosis of short circuit faults, even when they are not deep, to prevent their further development is crucial to ensuring the safe and reliable operation of batteries.

[0003] Currently, the main methods for battery fault diagnosis can be categorized as follows: 1) Model-based approach: First, a nonlinear model of the battery is constructed, and then the abnormalities in the estimated state are used to diagnose the fault. This approach depends on the accuracy of the constructed model and is easily affected by external interference, resulting in poor performance in practical applications.

[0004] 2) Signal processing-based methods: These methods utilize signal processing techniques, such as correlation coefficients and sample entropy, to process and analyze the collected battery data for fault diagnosis. However, this method suffers from low nonlinear fitting, leading to insufficient accuracy in fault diagnosis.

[0005] 3) Machine learning-based methods: These methods utilize the powerful nonlinear fitting capabilities of machine learning techniques such as neural networks to learn battery fault patterns from measured battery fault data, thereby performing fault diagnosis with high accuracy. However, this method suffers from a lack of high-quality fault data for training, resulting in poor model training performance.

[0006] Therefore, existing battery short-circuit fault diagnosis methods struggle to balance the accuracy of fault diagnosis with the demand for high-quality fault data, thus failing to ensure the safe and reliable operation of batteries. Summary of the Invention

[0007] To address the aforementioned issues, this disclosure proposes a battery short-circuit fault diagnosis method and system based on impedance spectrum. By selecting fault datasets for fault diagnosis models based on the importance of frequency characteristics, the information content of the fault datasets is improved, enabling accurate diagnosis of battery short-circuit faults even with only a small amount of battery fault data.

[0008] According to some embodiments, the present disclosure adopts the following technical solutions: A battery short-circuit fault diagnosis method based on impedance spectrum, comprising: Acquire impedance spectrum data of the battery to be diagnosed at the optimal sampling frequency band and perform preprocessing; The preprocessed impedance spectrum data is input into the trained fault diagnosis model to determine whether a short circuit fault has occurred and obtain the fault diagnosis result. The optimal sampling frequency band is constructed by using a fault diagnosis model to quantify the impact of impedance data at each sampling frequency on the fault diagnosis results, obtain the importance of frequency features, and construct the optimal sampling frequency band based on the importance of frequency features.

[0009] According to some embodiments, the present disclosure adopts the following technical solutions: A battery short-circuit fault diagnosis system based on impedance spectrum includes: The data acquisition module is configured to acquire impedance spectrum data of the battery to be diagnosed at the optimal sampling frequency band and perform preprocessing. The fault diagnosis module is configured to input the preprocessed impedance spectrum data into the trained fault diagnosis model, determine whether a short circuit fault has occurred, and obtain the fault diagnosis result. The optimal sampling frequency band is constructed by using a fault diagnosis model to quantify the impact of impedance data at each sampling frequency on the fault diagnosis results, obtain the importance of frequency features, and construct the optimal sampling frequency band based on the importance of frequency features.

[0010] According to some embodiments, the present disclosure adopts the following technical solutions: A computer program product includes a computer program that, when executed by a processor, implements the aforementioned battery short-circuit fault diagnosis method based on impedance spectrum.

[0011] According to some embodiments, the present disclosure adopts the following technical solutions: A non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the aforementioned battery short-circuit fault diagnosis method based on impedance spectrum.

[0012] According to some embodiments, the present disclosure adopts the following technical solutions: An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform the aforementioned battery short-circuit fault diagnosis method based on impedance spectrum.

[0013] Compared with the prior art, the beneficial effects of this disclosure are as follows: This invention combines the advantages of artificial intelligence and electrochemical impedance spectroscopy to propose a novel data-driven learning strategy for detecting short-circuit faults in batteries. Specific advantages include: (1) The method for short-circuit fault detection using impedance spectrum proposed in this invention utilizes the feature importance quantification function of TabPFN (Tabular Prior-data Fitted Network) to select the sampling frequency band impedance spectrum that has the greatest impact on fault detection. Fault detection can be completed when only the impedance spectrum of the battery to be diagnosed needs to be collected, which can effectively shorten the impedance spectrum acquisition time.

[0014] (2) The present invention uses impedance spectrum as fault detection data, which can fully reflect fault information. The TabPFN model used has significant performance advantages in processing small sample data. It can effectively deal with the problem that the fault is difficult to identify in the early stage of battery short circuit fault due to the weak changes in electrical and temperature signals, and meet the needs of accuracy, efficiency and real-time in actual application scenarios.

[0015] (3) The method developed in this invention has data-driven characteristics and does not involve complex battery electrochemical mechanisms. It can be conveniently used for fault diagnosis of various types of batteries without having to build different battery models under different conditions, as is the case with model-based methods. Attached Figure Description

[0016] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0017] Figure 1 This is a flowchart of the battery short-circuit fault diagnosis method in Example 1; Figure 2 This is a flowchart of Example 1, showing the detection of the battery to be diagnosed using a trained TabPFN model. Detailed Implementation The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0018] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0019] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0020] Example 1 One embodiment of this disclosure provides a battery short-circuit fault diagnosis method based on impedance spectrum, including: Step 1: Obtain the impedance spectrum data of the battery to be diagnosed at the optimal sampling frequency band and perform preprocessing; Step 2: Input the preprocessed impedance spectrum data into the trained fault diagnosis model to determine whether a short circuit fault has occurred and obtain the fault diagnosis result; The optimal sampling frequency band is constructed by using a fault diagnosis model to quantify the impact of impedance spectrum data at each sampling frequency on the fault diagnosis results, obtaining the importance of frequency features, and constructing the optimal sampling frequency band based on the importance of frequency features.

[0021] As one embodiment, this disclosure discloses a battery short-circuit fault diagnosis method based on impedance spectroscopy. First, an electrochemical workstation is used to collect impedance spectral data of batteries with different short-circuit faults at different states of charge. Then, TabPFN is used to quantify the importance of impedance spectra at each frequency for fault detection. The frequency band with the highest importance is then selected as the input feature for fault detection, and the TabPFN model is trained. Finally, the impedance spectrum of the selected frequency band of the battery to be diagnosed is collected and input into the TabPFN model to complete fault detection. Figure 1 As shown, the specific implementation process is as follows: S1: Select multiple batteries for short-circuit fault testing. Collect impedance spectrum data of normal batteries and batteries with different short-circuit fault levels at impedance sampling points corresponding to different sampling frequencies. The number of batteries n is selected according to the actual application. The main acquisition process of impedance spectrum is as follows: S11: Use a battery tester to charge or discharge the battery to a preset state of charge, with the selected state of charge being evenly distributed between 0% and 100%.

[0022] S12: After sufficient settling time, collect impedance spectrum data of normal batteries and impedance spectrum data of short-circuit fault batteries (with resistors of different resistance values ​​connected in parallel to the batteries).

[0023] S13: Before acquiring the impedance spectrum of the battery, determine, but not limited to, the frequency range and sampling density of the impedance spectrum. For a single impedance sampling point, the frequency, phase angle, and amplitude of the actual measured impedance must be recorded. The impedance spectrum data consists of impedances at different sampling frequencies, and the impedance is expressed by the formula:

[0024] In the formula, For impedance, Angular frequency, , Pi Sampling frequency, and These are the amplitudes of the excitation voltage and the response current, respectively. and These are the real and imaginary parts of the impedance, respectively.

[0025] S2: Preprocess and label the acquired impedance spectrum data to obtain the fault dataset. The specific steps are as follows: S21: Use denoising methods, including but not limited to median absolute deviation (MAD) denoising, to remove gross errors in the data, and replace missing values ​​with the average value.

[0026] S22: Normalize the impedance spectrum of the battery at various frequencies:

[0027] In the formula, X It is the raw data. It is normalized data. max(X) The maximum value in the original data. min(X) The minimum value in the original data.

[0028] S23: Mark the collected impedance spectrum of normal batteries as 0 and the impedance spectrum of faulty batteries as 1 to obtain a complete fault dataset.

[0029] S3: Using the complete fault dataset obtained above, initially train the TabPFN model, and use the feature importance quantification function of the TabPFN model to quantify the impact of impedance at each frequency on fault detection.

[0030] Among them, the TabPFN model, due to its pre-training mechanism, demonstrates superior performance compared to other algorithms on small-sample tabular data. This makes it well-suited for processing battery impedance spectral data with scarce samples due to difficulties and time-consuming data collection. The specific principle of the TabPFN model is as follows: TabPFN is a Transformer-based prior data fitting network (PFN). Based on the original PFN architecture, TabPFN primarily makes two architectural improvements to adapt to tabular data: The original PFN architecture uses a single multi-head self-attention module to compute attention between all training examples, as well as attention from validation examples to training examples. TabPFN replaces this with two modules that share weights: one module computes self-attention between training examples, and the other module computes cross-attention only from validation examples to training examples. This allows all examples to focus on themselves.

[0031] The input data has unequal numbers of dimensions k. PFN uses an encoder layer that accepts a fixed number of dimensions, while TabPFN amplifies the feature values ​​by padding the input dimensions with 0s to a fixed maximum dimension K. This allows Transformer to handle these datasets of different dimensions in a unified manner.

[0032] In this step, TabPFN, after pre-training and initial training, is used to quantify the impact of impedance spectra at various frequencies on fault detection.

[0033] During pre-training, the model is trained on a synthetic tabular dataset to minimize the difference between the predicted and true labels, thereby capturing generalizable patterns across different feature distributions. This process is typically performed offline by the developers.

[0034] Initial training is based on the complete fault dataset mentioned above. TabPFN performs contextual learning (ICL), which learns to make predictions using the sequence of labeled examples (x, f(x)) given in the input without further parameter updates. TabPFN accepts training and test samples as setpoint inputs and produces predictions for the entire test set in one forward pass.

[0035] Specifically, the fault dataset The TabPFN model is divided into training and testing sets to train it and search for the optimal distribution pattern until the classification results on the testing set meet the performance requirements.

[0036] After initial training is completed and a TabPFN model that meets the performance requirements is obtained, TabPFN is used to quantify the characteristic importance of impedances at each frequency.

[0037] TabPFN quantifies feature importance through SHAP (Shapley Additive Explanation). This method uses game theory principles to assign contribution values ​​to each feature and quantifies how each feature affects model predictions. The SHAP value of feature i, i.e., its importance, is calculated using the following formula:

[0038] in, It is a prediction function. It is the input to the prediction function. yes In the set of features, N is the set of all features, and S is the subset of features excluding feature i.

[0039] In this embodiment, f is the TabPFN model after preliminary training, i is the impedance at a single sampling frequency, and x is the impedance spectrum data of the selected frequency band. The feature importance of the impedance at a single sampling frequency is determined by the SHAP method.

[0040] S4: Based on the feature importance quantification results, select the impedance spectrum of the optimal frequency band for subsequent TabPFN model training. The specific method for selecting the optimal sampling frequency band is as follows: S41: Arrange the impedance spectrum sampling points from high to low according to the sampling frequency, and divide them into three segments with the same number of sampling points: high frequency, mid frequency, and low frequency. S42: Calculate the sum of the importance of all frequency features in each frequency band, select the frequency band with the largest sum of feature importance, and denot its sum of feature importance as... ; S43: Calculate the sum of the importance of all frequency features. , Will and Comparison, Based on the model training results Value adjustment, for example, if based on the current If the model trained using the optimal sampling frequency band does not achieve the required performance, then the value should be appropriately increased. The optimal sampling frequency band was reselected to meet the performance requirements of the model. S44: If If the frequency band is directly selected as the optimal sampling frequency band, then the other frequency band is selected as the optimal sampling frequency band. Based on this, increase the feature importance of adjacent frequencies until it is greater than This frequency band, along with the added adjacent frequencies, will be used as the optimal sampling frequency band. S45: Train the TabPFN model using a new fault dataset composed of the impedance spectrum of the optimal sampling frequency band and the corresponding labels until the required accuracy is achieved.

[0041] S5: As Figure 2 As shown, the impedance spectrum of the battery to be diagnosed in the optimal sampling frequency band selected in S4 is collected and used as the input of the trained TabPFN model to complete the short circuit fault detection.

[0042] The battery to be diagnosed is the same model as the battery provided in step S1 for training data. When the model output is 0, it means that the battery is normal. When the output is 1, it means that the battery has a short circuit fault.

[0043] Example 2 One embodiment of this disclosure provides a battery short-circuit fault diagnosis system based on impedance spectrum, comprising: The data acquisition module is configured to acquire impedance spectrum data of the battery to be diagnosed at the optimal sampling frequency band and perform preprocessing. The fault diagnosis module is configured to input the preprocessed impedance spectrum data into the trained fault diagnosis model, determine whether a short circuit fault has occurred, and obtain the fault diagnosis result. The optimal sampling frequency band is constructed by using a fault diagnosis model to quantify the impact of impedance data at each sampling frequency on the fault diagnosis results, obtain the importance of frequency features, and construct the optimal sampling frequency band based on the importance of frequency features.

[0044] Example 3 One embodiment of this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned battery short-circuit fault diagnosis method based on impedance spectrum.

[0045] Example 4 One embodiment of this disclosure provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the aforementioned battery short-circuit fault diagnosis method based on impedance spectrum.

[0046] Example 5 One embodiment of this disclosure provides an electronic device, including a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the battery short-circuit fault diagnosis method based on impedance spectrum.

[0047] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0048] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0049] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A battery short-circuit fault diagnosis method based on impedance spectroscopy, characterized in that, include: Acquire impedance spectrum data of the battery to be diagnosed at the optimal sampling frequency band and perform preprocessing; The preprocessed impedance spectrum data is input into the trained fault diagnosis model to determine whether a short circuit fault has occurred and obtain the fault diagnosis result. The optimal sampling frequency band is constructed by using a fault diagnosis model to quantify the impact of impedance data at each sampling frequency on the fault diagnosis results, obtain the importance of frequency features, and construct the optimal sampling frequency band based on the importance of frequency features. The fault diagnosis model uses tabular prior data to fit the TabPFN network, which includes two training sessions. The first time, a complete fault dataset was used for training, and the trained fault diagnosis model was used to construct the optimal sampling frequency band. The fault dataset after being filtered using the optimal sampling frequency band was used for training a second time. The trained fault diagnosis model was then used for the final battery short-circuit fault diagnosis. The importance of the frequency features is calculated using the fault diagnosis model after the first training, by calculating the contribution of impedance to fault diagnosis at each sampling frequency using the SHAP method. The specific steps for constructing the optimal sampling frequency band based on the importance of frequency features are as follows: The impedance sampling points are arranged from high to low according to the sampling frequency, and divided into three frequency bands with the same number of impedance sampling points. Calculate the sum of the importance of all frequency features in each frequency band, select the frequency band with the largest sum of frequency feature importance, and denot the sum of frequency feature importance as . ; Calculate the sum of the importance of all frequency features. , Will and Comparison, ,like If the frequency band is directly selected as the optimal sampling frequency band, then the other frequency band is selected as the optimal sampling frequency band. Based on this, increase the frequency characteristic importance of adjacent frequencies until it is greater than 1. This frequency band, along with the added adjacent frequencies, will be used as the optimal sampling frequency band.

2. The battery short-circuit fault diagnosis method based on impedance spectrum as described in claim 1, characterized in that, The impedance spectrum data consists of impedances at different sampling frequencies, and the impedance is expressed by the formula: in, For impedance, Angular frequency, , Pi Sampling frequency, and These are the amplitudes of the excitation voltage and the response current, respectively. and These are the real and imaginary parts of the impedance, respectively.

3. The battery short-circuit fault diagnosis method based on impedance spectrum as described in claim 1, characterized in that, The complete fault dataset is obtained by selecting multiple batteries for short-circuit fault testing, collecting impedance spectrum data of normal batteries and batteries with different short-circuit fault levels at impedance sampling points corresponding to different sampling frequencies, and marking whether a short-circuit fault has occurred.

4. A battery short-circuit fault diagnosis system based on impedance spectroscopy, characterized in that, The battery short-circuit fault diagnosis method based on impedance spectrum as described in any one of claims 1-3 includes: The data acquisition module is configured to acquire impedance spectrum data of the battery to be diagnosed at the optimal sampling frequency band and perform preprocessing. The fault diagnosis module is configured to input the preprocessed impedance spectrum data into the trained fault diagnosis model, determine whether a short circuit fault has occurred, and obtain the fault diagnosis result. The optimal sampling frequency band is constructed by using a fault diagnosis model to quantify the impact of impedance spectrum data at each sampling frequency on the fault diagnosis results, obtaining the importance of frequency features, and constructing the optimal sampling frequency band based on the importance of frequency features.

5. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the battery short-circuit fault diagnosis method based on impedance spectrum as described in any one of claims 1-3.

6. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement a battery short-circuit fault diagnosis method based on impedance spectrum as described in any one of claims 1-3.

7. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform an impedance spectrum-based battery short-circuit fault diagnosis method as described in any one of claims 1-3.

Images